The International Journal of Multimedia & Its Applications (IJMA) Vol.7, No.4/5, October 2015 THE TURING TEST FOR TELEPRESENCE Mathias Johanson Alkit Communications AB, Sallarängsbacken 2, Mölndal, Sweden ABSTRACT The quality of high-end videoconferencing systems has improved significantly over the last few years enabling a class of applications known as “telepresence” wherein the users engaged in a communication session experience a feeling of mutual presence in a shared virtual space. Telepresence systems have reached a maturity level that seriously challenges the old familiar truism that a face-to-face meeting is always better than a technology-mediated alternative. To explore the state of the art in telepresence technology and outline future opportunities, this paper proposes an optimality condition, expressed as a “Turing Test,” whereby the subjective experience of using a telepresence system is compared to the corresponding face-to-face situation. The requirements and challenges of designing a system passing such a Turing Test for telepresence are analyzed with respect to the limits of human perception, and the feasibility of achieving this goal with currently available or near future technology is discussed. KEYWORDS Telepresence, Quality of Experience, Videoconferencing, Virtual Reality, Augmented Reality 1. INTRODUCTION In 1950, Alan Turing published a famous paper entitled “Computing machinery and intelligence” wherein he considers the question “Can machines think?” [1]. Although the scientific value of the paper can be questioned, it has undoubtedly been a great source of inspiration for many researchers within the field of artificial intelligence (AI). In search of an answer to the question, Turing proposes a reformulation whereby the original problem is reduced to an investigation of whether computers can successfully engage in an “imitation game.” The game, as Turing describes it, is played by three participants: a man (A) a woman (B) and an interrogator (C) of any sex. The object of the game for the interrogator, who stays in a room apart from the other two, is to determine which of the other two is the man and which is the woman. The communication between the two rooms is by means of written notes only. The object of A in the game is to try to cause C to make the wrong identification, while B on the other hand should try to assist C making the right identification. The reformulated question posed by Turing is now, can a machine (i.e. a computer) be programmed to successfully play the part of A in the imitation game? Although the imitation game is expressly presented as a challenge for the interrogator to determine the sex of the other two players, it is usually interpreted as the problem of deciding which of two subjects a computer is and which is a human. Today, more than half a century since Turing posed his question; AI researchers devote little if any attention to the Turing Test as a measure of artificial intelligence. In fact, this was never the aim of Turing. His intention was to provide a tangible example to aid the discussion of the philosophy of artificial intelligence. However, the historical value of the test as an inspiration for AI researchers, computer scientists and technology developers to design systems and algorithms with powers comparable to human intelligence has been tremendous. In this sense, the Turing Test is often colloquially used to represent an ultimate goal of an AI system. DOI : 10.5121/ijma.2015.7501 1 The International Journal of Multimedia & Its Applications (IJMA) Vol.7, No.4/5, October 2015 In this paper, I propose a test inspired by the Turing Test for evaluating the quality of telepresence systems. In place of Turing’s question “Can machines think?” I pose the question “Can technology eliminate the effect of distance in interpersonal communication?”. More simply put, is it possible to design a telepresence system that delivers a service that is indistinguishable from a face-to-face meeting? In a way, this is quite the opposite of Turing’s test. In his imitation game, the subjects are allowed to communicate only by writing, so that the content of the conversation is in focus. In the telepresence test, we are on the contrary only interested in how the communication itself is experienced, not the subject matter of the conversation. In order to study the question of whether technology can mediate interpersonal communication completely transparently, and to provide some sort of way to assess the quality attainable with a telepresence system, the problem must be defined more accurately. I will begin by defining what a telepresence system is. 2. TELEPRESENCE The meaning of the term Telepresence in the context of this paper is the sensation of being physically present at the same location as another person (or any number of persons), although in reality the persons are at different locations. A Telepresence System is a set of technologies enabling this feeling of mutual presence by its users. Since vision and hearing are our main means of interpersonal communication, the key technologies of telepresence systems in this use of the term are typically concerned with audiovisual communication. A telepresence system can hence be seen as a high quality videoconferencing system. However, the term videoconferencing is heavily overloaded and can refer to almost any application of video-mediated communication. In order for a videoconferencing system to qualify as a telepresence system, there must be at least some ambition to present the appearance of the participants in a lifelike manner. A desktop videoconferencing system or a videophone application on a smartphone hence cannot be said to be a telepresence system. The definition of telepresence discussed so far is focused on the sensation of being together with other persons who are not physically present at the same location, and being able to communicate with them. Another use of the term is concerned with the feeling of being physically present at a remote location (regardless of other people), or to have an effect on the remote environment, through telerobotics or teleoperation. Indeed, this was the focus of cognitive scientist Marvin Minsky’s eponymous article in Omni Magazine in 1980, wherein the term was coined [2]. Today, the term Telepresence is most often used in the former meaning, i.e. in relation to high-end videoconferencing, partly due to the term being popularized as a marketing device by video- conferencing system vendors in the early 21st century. Telepresence in Minsky’s connotation of the term is nowadays usually considered to fall within the field of Virtual Reality (VR). 2.1. Virtual Reality The concepts of Telepresence and immersive VR are strongly interrelated. In an attempt to define VR in terms of human experience, rather than technological hardware, Jonathan Steuer [3] defines Telepresence as “…the experience of presence in an environment by means of a communication medium,” where Presence is defined (based on Gibson [4]) “…as the sense of being in an environment.” This is a generic definition covering both the communicative and immersive aspects of telepresence. The telepresence of interest here can be seen as a special case of the VR sense. For the rest of this paper, telepresence will refer to high end multimedia communication applications where the ambition is to give the users the illusion of being physically together in a shared space. 2 The International Journal of Multimedia & Its Applications (IJMA) Vol.7, No.4/5, October 2015 2.2. Immersion Closely related to both telepresence and VR is the concept of immersion. In an immersive VR system, the user experiences himself or herself immersed in a virtual, computer generated world. This is typically achieved through head-mounted display technology, or through large, all- surrounding wall projection systems. Similarly, in an immersive telepresence system, the user experiences the technology mediating the presence of other users as being immersed transparently in the physical space he or she occupies. Whereas VR systems are mainly concerned with synthetic representations of environments and users (referred to as avatars), video-based telepresence systems rely on video capture and display technologies, carefully integrated in the physical environments of the users. Hybrid systems, using a combination of video and computer generated 3D graphics have also been designed. These approaches are often referred to as augmented reality or mixed reality systems. Telepresence systems relevant for our present study can include elements of both synthetic virtual environments and video communication technology. 2.3. Multi-user and multi-point systems A salient feature of the telepresence systems of interest here is that they are by definition always multi-user systems. At a minimum two users, in a point-to-point communication session interconnecting two geographically separated sites, are required to qualify as a telepresence application. Telepresence sessions can also be truly multipoint, interconnecting more than two sites, and there can be more than one participant physically present at each site. It is furthermore required that all users — not just one or a subset of them — experience the sensation of telepresence. In a point-to-point two-user scenario, both users should experience the presence of the other user in a shared space. In that sense telepresence systems are by nature symmetrical. 2.4. Distance-spanning technology In the above definition of telepresence I have stipulated, somewhat vaguely, that the users should be at different locations. To design a practical test in the spirit of Turing, I will have to be a bit more precise on what I mean by different locations. Is it enough for a system to interconnect two rooms in the same building to be considered a telepresence system? From a practical standpoint, a telepresence system must span distances that take some non-negligible time to travel to be of interest.